nav emailalert searchbtn searchbox tablepage yinyongbenwen piczone journalimg journalInfo journalinfonormal searchdiv searchzone qikanlogo popupnotification paper paperNew
2026, 05, No.472 67-75
从海德格尔寻视理论看生成式人工智能的教育限度——基于GPT和DeepSeek的实证研究
基金项目(Foundation):
邮箱(Email):
DOI:
发布时间: 2026-05-10
出版时间: 2026-05-10
移动端阅读
摘要:

生成式人工智能(GAI)凭借强大的文本生成与逻辑推理能力,具有教育变革的潜力,但其离身认知的技术本质使变革具有一定限度。该研究选取“液体转移”“电路焊接”两项实验操作,对ChatGPT、 DeepSeek-R1等五种大语言模型进行半结构化访谈,以具身认知的4E特征为框架展开分析,尝试透视其背后的寻视世界,以划分GAI的发展阶段并框定GAI的教育限度。研究发现, GAI虽能在身体性、嵌入性、生成性层面接近人类回答,但始终无法突破延伸性局限,表现为操作者与工具的主客分离以及工具的“不上手”。基于海德格尔的寻视理论可发现,人类立基于“上手操劳”所展开的寻视世界, GAI则立基于由概率统计与符号演算所构建的语言世界,其“语言”是对人类语言痕迹的数学模拟,而非对“存在”的亲历。GAI的功能局限于“指导手册”,与人类教师的“手把手”“指导”有本质区别。研究启示教育者重视实践教育、强化元认知能力培养、重申生命教育价值,实现人与科技的协同发展。

Abstract:

Generative Artificial Intelligence(GAI), with its capabilities in text generation and reasoning, holds potential for educational transformation, yet its disembodied cognitive nature imposes limits on such change. This study selects two experimental operations “ Liquid Transfer” and “Circuit Soldering,” and conducts semi-structured interviews with five large language models, including ChatGPT and Deep SeekR1. Using the 4E framework of embodied cognition as an analytical lens, the study attempts to reveal the circumspective world behind these models and thereby delineate their developmental stages and pedagogical boundaries. Findings indicate that while GAI approximates human responses in the embodied, embedded, and enactive dimensions, it fails to overcome the limitation of extension, manifested in separation between operator and tool and the tool's state of being unready-to-hand. Drawing on Heidegger's theory of Umsicht, it becomes evident that the human circumspective world is grounded in engagement with ready-to-hand equipment, whereas GAI' s world is rooted in a linguistic realm constructed through probability statistics and symbolic calculus. Its “language” is a mathematical simulation of human linguistic traces, not a lived experience of Sein(Being). Consequently, GAI' s function is confined to that of an “instruction manual,” fundamentally distinct from the hands-on guidance of a human teacher. The study urges educators to prioritize practical education, strengthen metacognitive cultivation, and reaffirm the value of life education, thereby fostering synergistic human-technology development.

参考文献

[1]崔宇红,白帆等.Chat GPT在高等教育领域的应用、风险及应对[J].重庆理工大学学报(社会科学),2023,37(5):16-25.

[2]吴青,刘毓文.Chat GPT时代的高等教育应对:禁止还是变革[J].高校教育管理,2023,17(3):32-41.

[3]王佑镁,王旦等.“阿拉丁神灯”还是“潘多拉魔盒”:ChatGPT教育应用的潜能与风险[J].现代远程教育研究,2023,35(2):48-56.

[4]贾积有,张誉月.人工智能与教育:机遇、挑战与对策[J].北京大学教育评论,2023,21(1):49-61+188-189.

[5]荀渊.Chat GPT/生成式人工智能与高等教育的价值和使命[J].华东师范大学学报(教育科学版),2023,41(7):56-63.

[6]赵丽红,左敏等.人工智能时代高等教育教学的变革指向:培养高阶思维[J].北京师范大学学报(社会科学版),2023,(4):40-48.

[7]钟秉林,尚俊杰等.Chat GPT对教育的挑战(笔谈)[J].重庆高教研究,2023,11(3):3-25.

[8]张绒.生成式人工智能技术对教育领域的影响——关于Chat GPT的专访[J].电化教育研究,2023,44(2):5-14.

[9]李丽萍,卢晓东.实验教育中面向创新的具身认知发展价值——以化学学科为例[J].现代大学教育,2023,39(2):103-110.

[10]WEIZENBAUM J. ELIZA—a computer program for the study of natural language communication between man and machine[J].Communications of the ACM,1966,9(1):36-45.

[11]BROWN P F,DESOUZA P V,et al.Class-based n-gram models of natural language[J].Comput Linguist,1992,18(4):467-479.

[12]BENGIO Y,DUCHARME R,et al.A neural probabilistic language model[J].Journal of Machine Learning Research,2003,3(6):1137-1155.

[13]VASWANI A,SHAZEER N,et al.Attention is all you need[A].Advances in Neural Information Processing Systems 30[C].Long Beach:Curran Associates,Inc.,2017.5998-6008.

[14]CHRISTIANO P, LEIKE J, et al. Deep reinforcement learning from human preferences[A].Advances in Neural Information Processing Systems 30[C].Long Beach:Curran Associates,Inc.,2017.4299-4307.

[15]SEARLE J R.The Chinese room revisited[J].Behavioral and Brain Sciences,1982,5(2):345-348.

[16][德]海德格尔.陈嘉映,王庆节译.存在与时间[M].北京:商务印书馆,2018.

[17][英]波兰尼.张一兵译.认知与存在[M].南京:南京大学出版社,2017.

[18]胡立刚.论海德格尔“语言是存在的家”[J].齐齐哈尔大学学报(哲学社会科学版),2019,(4):36-38.

[19]SANCHES DE OLIVEIRA G.From something old to something new:functionalist lessons for the cognitive science of scientific creativity[J].Frontiers in Psychology,2022,12:750086.

[20]叶浩生.具身认知的原理与应用[M].北京:商务印书馆.2017.

[21]朱俊华,许璐瑶等.生成式人工智能如何赋能学生学习——基于大学生自我调节学习行为的实证研究[J].高等工程教育研究,2025,(2):66-72.

[22]沈书生,祝智庭.Chat GPT类产品:内在机制及其对学习评价的影响[J].中国远程教育,2023,43(4):8-15.

[23]王建虎,王静等.具身智能赋能教育的内涵、框架与研究路向[J].电化教育研究,2025,46(11):18-24.

[24]教育部.教育部关于印发《大中小学劳动教育指导纲要(试行)》的通知[J].中华人民共和国教育部公报,2020,(Z2):2-11.

[25]何卫平.解释学循环与美诺悖论——来自海德格尔思想的启示[J].学术月刊,2019,51(11):15-23+39.

[26]周洪宇,齐彦磊.教联网时代的生命教育:智能与生命的双和谐[J].现代教育管理,2020,(8):1-7.

(1)“中文房间理论”是美国哲学家约翰·希尔勒(John Searle)于1980年提出的思想实验。它设想一个人被置于一间房间内,不懂中文,但可以通过一套规则来回答中文问题。希尔勒认为,尽管这个人能够表现出对中文的理解,但他只是机械地按照规则操作,而非真正理解中文。这一理论对于人工智能是否能真正理解语言提出了质疑,并强调了“理解”与简单符号操作之间的区别。

(1)美诺悖论(Meno’s Paradox)由柏拉图在《美诺篇》中提出,核心质疑“人如何学习未知的事物”:若你已知某知识,便无需学习;若你全然不知,又无法意识到需要学习它,更无从探索——这一矛盾揭示了认知与无知之间的根本困境。

基本信息:

中图分类号:G434

引用信息:

[1]曾治,张誉月,卢晓东.从海德格尔寻视理论看生成式人工智能的教育限度——基于GPT和DeepSeek的实证研究[J].中国电化教育,2026,No.472(05):67-75.

发布时间:

2026-05-10

出版时间:

2026-05-10

检 索 高级检索